Wednesday night starting around 8:30 EDT (01:30 UTC) we will be upgrading our aging SQL hardware. Though we are fine on CPU and memory, we are out of storage space in our current Dell R710 generation servers that only have 8 drive bays. This is for the SQL cluster that houses the 335 databases for Stack Exchange Q&A sites as well as Careers…pretty much everything except Stack Overflow itself.
We will be dropping in new Dell R730xd boxes with the following specs:
- Processor: Dual Xeon E5-2667 v3 @3.2GHz
- Memory: 768GB (24x 32GB)
- Controller: PERC H730P
- Disk: 2x 2.5" 500GB SATA (OS - RAID 10), 1x Intel P3700 2TB NVMe PCIe SSD (Fast Data), and 24x 2.5" 1.2TB (10K SAS Spinny/Slow Data - RAID 10)
- Network: 2x 10Gb (via Intel X540 NDC in LACP)
We expect this maintenance to last a few hours with user-facing downtime being only two 10-15 second blips as well as two 10 minute read-only periods where we test each of the 2 new servers.
This maintenance will be repeated soon in our Oregon data center to upgrade the cluster node there, but that window should have no user-facing impact.